Multispectral object detection using 3D models cascade and region mutual information

The visual appearance of an object under natural lighting depends on its surface reflectance. However, there are situations when objects are to be viewed under non-visible light. Under such circumstances where reflectance information is not available, it is infeasible to use existing appearance-base...

Full description

Saved in:
Bibliographic Details
Main Author: Pong, Hon Keat.
Other Authors: Cham, Tat Jen
Format: Theses and Dissertations
Language:English
Published: 2008
Subjects:
Online Access:http://hdl.handle.net/10356/13604
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:The visual appearance of an object under natural lighting depends on its surface reflectance. However, there are situations when objects are to be viewed under non-visible light. Under such circumstances where reflectance information is not available, it is infeasible to use existing appearance-based object models which are based on surface reflectance. This thesis presents methods for detecting object instances within images, with efforts directed at developing methods that do not assume prior knowledge about reflectance information. We adopt the model based approach, in which a 3D model is used for detection by aligning it to image. An algorithm to speed up detection is also presented.